political intervention in economic...
TRANSCRIPT
SERIE DOCUMENTOS DE TRABAJO No. 83
Junio 2010
POLITICAL INTERVENTION IN ECONOMIC ACTIVITY
Enrique Gilles
Political intervention in economic activity
Enrique Gilles∗
Abstract
This paper proposes a political economy explanation of bailouts to declining
industries. A model of probabilistic voting is developed, in which two candidates
compete for the vote of two groups of the society through tactical redistribution.
We allow politicians to have core support groups they understand better, this im-
plies politicians are more or less effective to deliver favors to some groups. This
setting is suited to reproduce pork barrels or machine politics and patronage. We
use this model to illustrate the case of an economy with both an efficient industry
and a declining one, in which workers elect their government. We present the con-
ditions under which the political process ends up with the lagged-behind industry
being allowed to survive.
JEL Codes: D72, D78.
Keywords: probabilistic voting, redistribution, survival of inefficient firms, pork
barrels.
∗I am grateful to Timothy Kehoe, Igor Livshits, Darıo Maldonado, Oskar Nupia, Franck Portier,
Vıctor Rıos-Rull, Monica Vargas and seminar participants at the XII Workshop on Dynamics Macroe-
conomics at Vigo, LACEA 2007, and Universidad del Rosario for helpful comments. All remain-
ing errors are my own. Universidad del Rosario, Bogota. E-mail: [email protected].
The findings, recommendations, interpretations and conclusions expressed in this paper are those of
the author and not necessarily reflect the view of the Department of Economics at Universidad del
Rosario.
1
1 Introduction
One of the first lessons of Economics refers to an infinity of agents taking economic
decisions based on all the relevant information: this is our first approach to the perfect
competition hypothesis. In the particular case of firms, the lesson continues, we learn
they take prices as given, there are many of them, that their only goal is to maximize
their profits, and that they are free to enter and leave markets. We soon realize these
conditions are too strict, and that it is hard to find a perfectly competitive market. We
then start learning about monopolies, oligopolies, and about barriers to entry: every-
thing becomes clearer. Perfect competition also implies that only the most efficient firms
can survive the harshness of markets, while the others are condemned to dissappear.
This paper is about why those other firms are sometimes not allowed to fail. More
precisely, we analyze government intervention that prevents firms to die: this constitutes
another departure from the perfect competition hypothesis. The political intervention
into economic activity is closely related to the soft budget constraint syndrome, i.e. an
economic disease –deeply rooted in former socialist and transitional countries– that is
characterized by persistent rescues of failing institutions by a support institution. The
syndrome is fully at work when this intervention creates in turn expectations of future
bailouts1.
The softening of budget constraints is not exclusive of transition economies. Take
the case of the US automobile industry: there is a consensus around the fact that their
financial troubles are closely related to the persistence of poor business practices (lack of
innovation mainly) together with a unionized workforce. These problems have been aris-
ing since the late seventies, and their partial resolution has often implied government’s
transfers. This support has crystallized under different forms: subsidized credit, bridge
loans, and backing of warranties. Thirty years later, we witness the end of an era in
the American automobile industry, and there is enough evidence to assert that repeated
bailouts have had a negative impact on the incentives to innovate of these firms. In
turn, successive governments, either captured by interest groups or invoking the need to
protect employment, have all shown some degree of willingness to intervene.
1For an extensive review on the literature of the soft budget constraint, see Kornai et al (2003).
2
Hence, if we are to understand the reasons of the persistence of government support
and the survival of an inefficient industry, we cannot avoid discussing political factors
such as rent seeking, lobbying activities and the political power of concerned groups.
Politicians take economic decisions, and often these decisions do not pursue economic
efficiency but they reflect the political power of interest groups.
We present a model of probabilistic voting to address this issue of political interven-
tion in economic activity. We identify the conditions under which two interest groups
may benefit or suffer from redistribution schemes as a result of the political game be-
tween two office-motivated candidates. These candidates are characterized by different
abilities to deliver political favors to interest groups.
Dixit and Londregan (1996) analyze the main characteristics that special interest
groups may have in order to be benefitted by redistribution in electoral contests. They
identify those traits in two settings: if parties have no differences in their ability to
deliver favors, the benefits of redistribution will be mainly targeted to swing voters.
On the other hand, with different abilities to transfers, parties will favor their core
support groups. These abilities are synthetized in the leaky bucket assumption: there
are deadweight losses in the redistribution activity, that depend on parties’ proximity to
voters. When parties have core support groups they understand better, they can more
easily target subsidies to them in exchange of votes. It is then a rationale for “pork
barrel” politics, “machine politics”, or “patronage”, that is, spending that is intended
to benefit constituents of a politician in return for their political support, either in the
form of campaign contributions or votes.
In a related work, Dixit and Londregan (1995) address the issue of redistributive
politics and economic efficiency. Using a probabilistic model they show that workers in
a lagged-behind industry benefitting from subsidies may have no incentives to relocate
to others, more efficient industries. Then inefficient industries may be totally locked
in: their workers know that if they move nothing guarantees they will continue to be
recipients of political redistribution. So even if the decision to move would imply a
higher labor income, this advantage may vanish when political considerations are taken.
Robinson and Torvik (2005) propose a political economy model to explain the ex-
istence and persistence of soft budget constraints. Their starting argument is that an
3
incumbent politician may be willing to launch projects known to be inefficient, pro-
vided only him is able to refinance them tomorrow, i.e. after elections. This creates
incentives for some voters to vote for him, since it is the only way their projects will
be bailed out. The soft budget constraint syndrome is not viewed as in the standard
way, i.e. a problem, but as an opportunity for politicians to get electoral support. The
approach followed in the paper is based on a probabilistic model and using the leaky
bucket assumption, but in a simplified way than the one we develop here.
In this paper we take the leaky bucket assumption of Dixit and Londregan (1996)
model and, by focusing on two special interest groups, a given utility function and a
distribution of ideology among groups, we are able to obtain a closed form solution for
the redistributive scheme of each candidate. Furthermore, we show that under different
candidates’ abilities to redistribute, parties platforms differ. Usually, the equilibrium in
probabilistic voting models entails both parties proposing the same policies. However,
we should recognize that this is hardly found in real life politics (e.g. conservatives
are more pro-market whereas liberals are more interventionists, and this surely implies
different policies). Our model is capable to reproduce this divergence, but in order
to know which is the implemented policy, it is relevant to know who is the winner of
elections in a given political equilibrium. We derive conditions to identify the winner.
We then setup an example in which a lagged-behind industry is allowed to survive
given that its workers are key in electoral terms to one of the politicians. In this sense, we
offer a model that sheds some light, from a political economy perspective, in explaining
the occurrence and persistence of bailouts.
The paper is presented as follows: in section 2 we present the political equilibrium
of the model. Section 3 presents the economic model. Section 4 concludes.
2 Political equilibrium
Consider an economy in which the society is represented by two groups, indexed by
i = {1, 2}. Group 1 has N1 members whereas the number of individuals in group 2 is
N2. Since we are not concerned with the decision of whether to vote, we assume each
group member actually votes, so we are calling them voters (and also, for the purposes
4
of the next section, workers). The total vote is thus equal to the total population
N = N1 +N2.
There are 2 political parties (also called candidates) denoted by k = {A,B}. These
candidates compete to attract voters in order to maximise their total vote, by proposing
transfers to each group. Their motivation is only to be in power (from which they may
extract exogenous rents) and they do not have partisan preferences. However, they
do have some observable characteristics to voters, namely an issue position, ideology
or even popularity. In any case, these attributes are considered as fixed so that they
cannot be accomodated during a given electoral process. Parties platforms consist in
a redistributive scheme to group members, which is implemented with certainty if the
party wins the elections. This last assumption on enforceability will allow us to name
parties indistinctly as “candidate k” or, in some contexts, simply “the government”.
Platforms are denoted by Tik, that is the amount of redistribution from party k to each
of group i members.
Leaky bucket assumption: the technology to collect taxes and pay subsidies implies
that there are leakages between what the government collects (Tik) and what agents
receive (tik):
tik =
(1 + γik)Tik if Tik < 0
(1 − θik)Tik if Tik > 0(1)
Where the (positive) parameters γik and θik indicate the ability party k has to tax or
transfer money to group i, respectively. Note these abilities may be asyimmetric, i.e. a
candidate may be skilled in transfering to some particular group and at the same time
have less ability to tax the other group, or viceversa. Parameter values close to zero
mean a close relationship between politicians and group members, while high values
indicate the opposite. Parties have the following budget constraint2:
∑
i
NiTik = 0 (2)
The level of income ωik of an individual in group i when candidate k is on office, is
2We could have a positive amount available for redistribution, at a higher cost in terms of algebra.
5
composed by wages and transfers:
ωik = wi + tik (3)
Wages are group-specific and they are set as the marginal productivity of labor. Indi-
viduals derive utility from income, their indirect utility function is given by:
Ui(ωik) = σ lnωik, 0 < σ < 1
The sequence of events in this game is the following: first, parties choose their
platforms {Tik}, elections take place, and finally the winner implements its policy.
In deciding their vote, workers care not only about economics but also about ideology.
This means they are sensitive to transfers but they also have an issue position. We rule
out the possibility to interpret group membership as deriving from ideological aspects,
instead, we allow for political diversity within each group. We assume that ideology is
continously distributed among voters of each group, such that a voter located at X has
an ideological preference X for candidate B.
The voting decision is therefore the following: a worker of group i votes for candidate
A if the indirect utility she gets should A win more than compensates the utility she
would get if B wins, taking into account her ideological preferences over B, i.e.
vote for A ⇔ Ui(ωiA) > Ui(ωiB) +X.
Given the distribution of ideology among voters, we can define the marginal voter in
group i, that is, the one who is just indifferent between voting for candidate A and
candidate B: this indifferent voter is characterized by having an ideology bias towards
candidate B that we denote Xi, defined then as
Xi ≡ Ui(ωiA) − Ui(ωiB).
All individuals in group i characterized by an ideology level X < Xi vote for candidate
A. Taking into account our specific indirect utility function, this cutpoint is:
Xi ≡ σ lnωiA − σ lnωiB = σ ln
(
ωiA
ωiB
)
.
6
We assume we know the distribution of the exponential of X,
eX ≡ x ∼ U
[
1 −1
2δi, 1 +
1
2δi
]
where δi is the probability density function of x in group i. Then the ideology cutpoint
is
xi =
(
ωiA
ωiB
)σ
. (4)
Letting Φi(x) be the cumulative distribution function of x in group i, the proportion of
members of group i voting for party A is given by:
Probi(x ≤ xi) = Φi(xi) = δi(xi − 1) +1
2(5)
whereas the proportion of group i individuals voting for B is
Probi(x ≥ xi) = 1 − Φi(xi) =1
2− δi(xi − 1) (6)
we have∂Φi(xi)
∂xi= φi(xi) ≡ δi
Φi(1) =1
2and E(x) = 1.
From the above facts, when parties choose their platforms the thresholds xi are deter-
mined, and we can express the total vote for each party as
VA =i=2∑
i=1
NiΦi(xi), (7)
and
VB =i=2∑
i=1
Ni [1 − Φi(xi)] = N − VA. (8)
2.1 Which group will be benefitted by redistribution?
Both parties choose their platforms {Tik} so as to maximize their total vote Vk. As the
budget constraint –equation (2)– indicates, if parties make positive transfers to a group,
the other group is necessarily being taxed. The problem at this point in the analysis
7
is to find under which conditions a given group will be taxed while the other gets a
subsidy. Which are the main features that an interest group must have in order to profit
from redistribution? Let us define the following variables that help deal with the leaky
bucket parameters:
when party k taxes group 1 and subsidizes group 2, we use: ψk =1 − θ2k
1 + γ1k
when party k taxes group 2 and subsidizes group 1, we use: ǫk =1 − θ1k
1 + γ2k
See that the maximum value of ψk and of ǫk is one, corresponding to the case θik = γik =
0, that is, when there are no leakages between the government and the groups. Thus,
the closer ψk and ǫk to unity, the more ability party k has in the redistributive activity.
Proposition 1 Groups benefitted from redistribution
• Party A taxes group 1 to subsidize group 2 if δ2δ1
w1
w2>
ψσ
B
ψ1+σ
A
• Party A taxes group 2 to subsidize group 1 if δ2δ1
w1
w2<
ǫ1+σ
A
ǫσB
• Party B taxes group 1 to subsidize group 2 if δ2δ1
w1
w2>
ψσ−1
B
ψσ
A
• Party B taxes group 2 to subsidize group 1 if δ2δ1
w1
w2< ǫσAǫ
1−σB
Proof. See the Appendix.
We show these results in Figure 1. Group i will be benefitted the lower its wage wi
and the higher the density δi compared to the respective values of the other group.
In other words, parties will tax the group with higher pre-tax income, and they will
offer subsidies to groups with a high density. Note the uniform distribution assumption
implies a constant density and this simplifies the analysis. The condition actually says
that benefitted groups have a high density at the indifferent voter, i.e. a high φi(xi).
Therefore, parties will favor groups with a higher proportion of those marginal voters,
characterized by a higher willingness to trade ideology for transfers (i.e. more center
in the political spectrum). Another important result is that the number of members
of a group has no influence on the determination of the favored group. This means
8
that even a small group may be benefitted. Note that for some levels of wages and
densities, parties do not propose any transfer. Finally, we also show in the Appendix
that strategies are characterized by both parties taxing or subsidizing the same groups,
i.e. we cannot find a situation in which, say, party A taxes group 1 while party B taxes
group 2.
δ2δ1
w1
w2
ǫA
(
ǫAǫB
)σ
A taxes group 2
1
ψA
(
ψB
ψA
)σ
A taxes group 1
δ2δ1
w1
w2
ǫB
(
ǫAǫB
)σ
B taxes group 2
1
ψB
(
ψB
ψA
)σ
B taxes group 1
Figure 1: Groups benefitted from redistribution
A sufficient condition under which both parties tax group 1 and subsidize group 2 is
δ2
δ1
w1
w2
>
(
ψB
ψA
)σ
max
{
1
ψA,
1
ψB
}
(9)
i.e. if ψA < ψB and if party A taxes group 1 to subsidize group 2, then party B do the
same. If ψA > ψB the opposite holds.
2.2 The political game
Let us consider hereafter the case in which both parties decide to tax group 1, using
these proceeds to subsidize group 2. As seen above, this corresponds to a situation in
which group 1 wage is high with respect to the wage in group 2 and when δ2 is high
compared to δ1. Therefore we use the parameters embedded in ψk instead of those of ǫk.
Each party wants to maximize its total vote, Vk, by choosing its platform {Tik}, given
the platform the other candidate has selected. Note that, since parties have a budget
constraint, the problem is unidimensional: once T2k has been chosen, T1k is determined.
Choosing platforms determines the income levels ωik that group members will obtain
under each candidate, see equation (3). Then the thresholds xi –which are the main
9
inputs of the objective (vote) functions– are obtained. Finally, we get the total vote for
each party.
The problem of Party A is then the following:
max{TiA} VA =∑i=2
i=1NiΦi(xi)
s.t.∑i=2
i=1NiTiA = 0 BCA
ω1A = w1 + t1A > 0 LL1A
ω2A = w2 + t2A > 0 LL2A
Party A seeks to maximize its total vote subject to the budget constraint (BCA) and
two limited liability constraints (LLiA) to prevent agents get a non-positive income.
Substituting the expressions for xi, given by equation (4) and for Φi(xi) given by equation
(5), using the constraint BCA to eliminate T1A, the problem may be specified as
max{T2A} VA = N2
+N1δ1
[
(ω1A
ω1B)σ − 1
]
+N2δ2
[
(ω2A
ω2B)σ − 1
]
s.t. ω1A > 0 LL1A
ω2A > 0 LL2A
The first order conditions of this problem yield the best response function of party A
with respect to party B. We can express it as
χA =
[
δ1
δ2
1
ψA
]1
1−σ
[
1
χB
]σ
1−σ
(10)
where, for convenience, we use the variable χk defined as the following ratio:
χk =ω1k
ω2k
, (11)
and recall the levels of after-transfer income write
ω1k = w1 − (1 + γ1k)N2
N1
T2k
ω2k = w2 + (1 − θ2k)T2k.
The important thing to keep from this is each χk defines a unique value of T2k,
T2k =N1 (w1 − w2χk)
(1 − θ2k)N1χk + (1 + γ1k)N2
, (12)
10
with low values of χk corresponding to high values of the transfer to group 2 members,
T2k. Moreover, through the budget constraint, χk also defines a unique T1k.
Similarly, party B’s problem is to maximize its total vote, which, following similar
steps as done for A results in the following problem depending on T2B:
max{T2B} VB = N2−N1δ1
[
(ω1A
ω1B)σ − 1
]
−N2δ2
[
(ω2A
ω2B)σ − 1
]
s.t. ω1B = w1 + t1B > 0 LL1B
ω2B = w2 + t2B > 0 LL2B
The resulting best response function is given by:
χB =
[
δ1
δ2
1
ψB
]1
1+σ
χσ
1+σ
A (13)
Behavior of best response functions. Figure 2 shows the best response functions of
each party, together with the effects of a decrease in parties’ abilities to transfer funds.
These lower levels of abilities are modelled through an increase of the parameters θ2A
and θ2B (or also by an increase in the γ1k parameters). The bold lines represent the new
position of the best response functions. For party A, the new best response functions lies
to the north-east of the original one, while for party B it lies to the north-west. Since the
leakage occurred between the government and its beneficiaries diminishes the political
power of the former to attract voters, candidates are better off with low parameter
values. From this graphical analysis it follows that party A will prefer low leaky-bucket
parameter values such that its best response function lies the closer as possible to the
origin. Similarly, party B will prefer a best response function as closer to the horizontal
axis as possible, corresponding to low values of the leaky-bucket parameters.
2.3 Equilibrium
The Nash equilibrium of this game is found at the intersection of both best response
functions, equations (10) and (13):
χ∗A(χ∗
B) = χ∗B(χ∗
A)
11
ω1B
ω2B
ω1A
ω2A
χB(χA)
χA(χB)
Figure 2: Effects of a decrease in parties’ abilities to transfer
where starred variables denote equilibrium values. This equilibrium is then
χ∗A =
δ1
δ2
[
1
ψA
]1+σ
ψσB, (14)
χ∗B =
δ1
δ2
[
1
ψA
]σ [1
ψB
]1−σ
. (15)
Once the equilibrium values for these variables obtained, recalling equation (11), we
immediately get the corresponding equilibrium transfers T ∗2k,
T ∗2A =
N1
1−θ2A
[
δ2ψ1+σA w1 − δ1ψ
σBw2
]
N1δ1ψσB +N2δ2ψ
σA
(16)
T ∗2B =
N1
1−θ2B[δ2ψBψ
σAw1 − δ1ψ
σBw2]
N1δ1ψσB +N2δ2ψ
σA
(17)
Figure 3 shows both candidates’ best response functions and the resulting Nash
equilibrium, for the case in which party A is more efficient in redistribution than B, i.e.
for ψA > ψB.
We may next obtain the equilibrium levels of after-transfers income, ωik:
12
χB
χA
45◦
χB(χA)
χA(χB)
χ∗A
χ∗B
Figure 3: Nash Equilibrium – party A more efficient in redistribution
ω∗1A =
δ1ψ−1
A ψσBN1δ1ψ
σB +N2δ2ψ
σA
(N1ψAw1 +N2w2) (18)
ω∗2A =
δ2ψσA
N1δ1ψσB +N2δ2ψ
σA
(N1ψAw1 +N2w2) (19)
ω∗1B =
δ1ψσ−1
B
N1δ1ψσB +N2δ2ψ
σA
(N1ψBw1 +N2w2) (20)
ω∗2B =
δ2ψσA
N1δ1ψσB +N2δ2ψ
σA
(N1ψBw1 +N2w2) . (21)
It is interesting to remark some facts about the equilibrium levels of income. To do that,
let us first consider the case in which there are no leakages in redistribution such that
ψk = 1. The after-transfers income of an individual of group i may be written as follows:
ω∗ik =
δi∑
iNiδiY
13
where Y stands for the total income3 of the economy, that is: Y = N1w1 +N2w2. The
after-transfers income an individual of group i gets from party k is then some fraction
of the total income. This result can be interpreted as if the government collected all the
income to then redistribute it to group members according to some weights, and these
weights reflect the political power of a given group. Group i’s members will be benefitted
the higher their density δi, that is, the more centered the distribution of ideology among
voters. We have already mentioned a similar intuition when showing the features of
beneffited groups by redistribution. Also, group size plays no direct role on determining
the political power of a given group, but it has an indirect effect through the total income
available for redistribution in the economy. Moreover, the pre-transfers level of income
(i.e. wages) of an individual is almost irrelevant to determine her after-transfers income.
Things get less straightforward when we introduce the leaky bucket parameters.
However, we can highlight a similar intuition: the total income available to candidate k
for redistribution is now
Yk = N1ψkw1 +N2w2 (22)
this can be interpreted as the maximum amount of income candidate k has in order to
redistribute from group 1 to group 2. Indeed, suppose group 1 is fully taxed, then the
proceeds candidate k gets from taxing this group are
N1w1
1 + γ1k
that is lower than N1w1 due to the leakages in the way from group 1 to party k. This
amount would then serve to make transfers to the other group. Group 2 members would
receive, after the leaky-bucket losses
N1w1
1 + γ1k
(1 − θ2k) ≡ N1ψkw1
which, in addition to their own income N2w2 makes the total after-transfers income
available to candidate k for redistribution (Yk), for the case in which group 1 is taxed
to subsidize group 2. In turn, the weights that determine the allocation of that income
3In the absence of leakages and of any extra budget available for redistribution, the distinction
between before and after transfers is not relevant.
14
to the different groups take into account now the effects of the leakages accruing in
redistribution and, despite being not as much clear as for the no-leakages case, preserve
the same qualitative insights: benefitted groups are those with a higher density δi and
a better proximity to candidate k, i.e. a higher ψk.
Concerning the levels of the final income of each group member compared to their
pre-transfers levels, we have the following result.
Proposition 2 Groups with higher pre-transfer income may get a lower after-transfer
income, i.e. after the political process.
Proof. Let us consider the case in which group 2 members get a higher income after the
political process, i.e. ω∗2A > ω∗
1A despite having a lower wage than group 1 members.
The condition to have this is obtained from the equilibrium values of final income given
by equations (18) and (19):
ω∗2A > ω∗
1A ⇔ 1 >δ1
δ2
ψσBψ1+σA
We have now to check that this condition satisfies the requirements we have imposed to
have party A taxing group 1 to subsidize group 2. Recalling Proposition 1, we need
w1
w2
>δ1
δ2
ψσBψ1+σA
Therefore it may be the case that
w1
w2
> 1 >δ1
δ2
ψσBψ1+σA
.
The procedure if we consider party B transfers is analogous. 2
The political characteristics of groups determine which group will be taxed and which
will be subsidized. This result goes beyond that and better identifies winners and losers
from the electoral process: in the example, the high-wage agents (i.e. group 1 mem-
bers) not only subsidize the low-wage ones, but they also end up with a lower income,
after transfers are made. This situation may arrive if the poor are a more ideology-
concentrated group than the rich (i.e. if they belong to the center of the ideological
spectrum).
15
With equations (18) to (21), we finally determine the equilibrium values of the cut-
points x1 and x2,
x∗1 =
(
ω∗1A
ω∗1B
)σ
=
(
ψB
ψA
YA
YB
)σ
x∗2 =
(
ω∗2A
ω∗2B
)σ
=
(
YA
YB
)σ
It immediately follows that the following relationship holds
x∗1 =
(
ψB
ψA
)σ
x∗2 (23)
16
Analysis of the equilibrium:
Proposition 3 Under identical candidates’ abilities to transfer, i.e. θ2A = θ2B and
γ1A = γ1B we find the standard political equilibrium result in which parties’ platforms
are identical.
Proof. When θ2A = θ2B and γ1A = γ1B we have ψA = ψB = ψ. We first insert ψ into the
equilibrium values of χA and χB given respectively by equations (14) and (15). At this
point we have χ∗A = χ∗
B = χ∗. Then, the result follows from taking into account that
each χk defines a unique T2k (and hence a unique T1k) as equation (12) shows. 2
This case reproduces the convergence result of most political economy models, while
in this setting it is just a particular case.4 See Figure 4. Instead, if leaky bucket
parameters are such that ψA > ψB we lose platform convergence and the equilibrium is
found above the 45◦ line, as the example of Figure 3 showed. In such a setting party A
is able to extract more funds from group 1 and to subsidize more efficiently group 2 than
party B. The opposite situations emerges whenever ψB > ψA, in which the equilibrium
lies below the 45◦ line.
2.4 Election’s result
As seen above, given the different abilities parties have to redistribute to interest groups,
their plaftorms will in general differ. In order to determine which is the implemented
platform, we need to know who is the winner of the election. Recalling that every agent
of the economy votes, we propose the following result:
Proposition 4 In the case both parties tax group 1 to subsidize group 2, the party with
the highest ability to transfer to groups (i.e. with the highest ψk) wins the elections. If
both parties have the same ability to transfer (i.e. ψA = ψB) there is a tie in elections.
4There is also the situation in which θ2A 6= θ2B and γ1A 6= γ1B but the parameter levels still imply
ψA = ψB : Party platforms here differ but we have χ∗
A= χ∗
B= χ∗ as before.
17
ω1B
ω2B
ω1A
ω2A
45◦
χB(χA)
χA(χB)
χ∗
χ∗
Figure 4: Parties are equally efficient in redistribution
Proof. See the appendix.
3 The Economic model
In this section we propose an application of the political model introduced in the pre-
vious section. We now turn to the description of the economy. Groups 1 and 2 become
here the workers of two sectors or industries of this economy: industry 1 and industry 2.
We can think on sector 1 as being the new, highly productive industry. Sector 2 instead
is the old, low productive and declining one. Workers have in principle the possibility to
migrate from one sector to the other. However, we assume that in the short run there
are switching costs sufficiently high so as to impede workers to do that. Hence, there is
no labor mobility across sectors.
Sector 1 description The technology of sector 1 is represented by the following
production function
Y1 = F1(N1) = A1N1, A1 > 0,
where A1 is the marginal productivity of labor in sector 1. Wages in this industry, de-
18
noted by w1, are set as the marginal productivity of labor. The after–transfers income
ω1k of an individual working in this industry is given by this wage and the transfer she
pays to the government in case candidate k is in office
ω1A = w1 + t1A = w1 − (1 + γ1A)N2
N1
T2A
ω1B = w1 + t1B = w1 − (1 + γ1B)N2
N1
T2B
In sum, a little changes with respect to the canonical model presented in the preceding
section.
Sector 2 description The production function is
Y2 = F2(N2) = A2N2, A2 > 0.
The coefficient A2 is the marginal productivity of labor, we assume A2 < A1.
Assumption 1 Sector 2 is represented by a firm.
There is an additional agent in the model, the firm of industry/sector 2. We assume
that it is a separate entity, that is, it does not represent any of the other agents of the
economy, and we further assume it does not have any kind of political power. There
is no possibility that it may influence the political process, it undertakes no lobbying
activities, and therefore it plays a neutral role in the elaboration of parties’ platforms
and in determining the outcome of elections.
Assumption 2 Transfers to group 2 workers may be channelled through sector 2.
Sector 2 revenues are composed not only by its production but also by transfers t2k from
the government/candidate k. We depart here from the standard model presented before.
Thus, candidates’ platform consists in a transfer T2k made to sector 2 as a whole. The
leaky bucket assumption still plays a role, though: From this transfer, only an amount
t2k make it to sector 2, and this amount is then splitted between workers and the firm.
We thus have a “payroll subsidy”. Let us then define sector 2’s gross income as
GI2k = Y2 +N2t2k = N2(A2 + t2k),
19
Assumption 3 Income sharing between sector 2 workers and the firm is determined
according to the Generalized Nash Bargaining Solution.
We assume the firm has bargaining power µ. The agreement payoffs are given by:
firm : GI2k −N2ω2k
workers : N2ω2k
whereas disagreement payoffs are:
firm : 0
workers : 0
i.e. should no agreement take place both the firm and sector 2 workers get nothing. The
Nash Bargaining solution is given by the value of ω2k that solves the following problem:
max (GI2k −N2ω2k)µ (N2ω2k)
1−µ
{ω2k}
The solution is
N2ω2k = (1 − µ)N2(A2 + t2k) = (1 − µ)GI2k
i.e. a proportion (1 − µ) of gross income goes to workers, while the remaining µ to the
firm. Note we can still decompose worker’s income as a part of wages w2 = (1 − µ)A2
and a part of transfers (1−µ)t2k. Therefore, the transfer effectively received by workers
passes through two filters: the loss accruing between the government and the sector
due to the leaky bucket, and the sharing rule applied in sector 2, which depends on the
bargaining power of agents.
(1 − µ)t2k = (1 − µ)(1 − θ2k)T2k ≡ (1 − θ2k)T2k
where we introduce the parameter θ2k = 1 − (1 − µ)(1 − θ2k), which θ2k > θ2k > 0
whenever µ > 0. Accordingly we now write –for this case in which both parties tax
sector 1 to subsidize sector 2–
ψk =1 − θ2k
1 + γ1k
= (1 − µ)(1 − θ2k)
1 + γ1k
= (1 − µ)ψk for k = A,B.
20
Assumption 4 Sector 2 can be made operational only after a fixed cost F is invested.
After the bargaining process, the firm has to invest a fixed quantity F in order that the
sector can be working. The payoff the firm realizes is a level of profits Π2k that writes
Π2k = µGI2k − F = µN2(A2 + t2k) − F. (24)
It follows that there exists a level of transfers for which the firm makes zero benefits,
T2k =F − µN2A2
(1 − θ2k)µN2
. (25)
We now incorporate an extra feature to the model. As equation (25) indicates, the
firm’s profits are positive whenever transfers are greater than T2k. In such a case the
analysis is similar to the one introduced in section 2, properly adjusted to introduce the
modifications of this section. When transfers are lower than this cutpoint, we are in the
“constrained region” and the setting changes.
Assumption 5 The firm requires non-negative profits to operate Sector 2.
Note that sector 2 profitability heavily depends on government’s transfers. Indeed,
according to equation (24), in the absence of transfers firm’s benefits become:
Π2 = µN2A2 − F,
these profits will be negative in the relevant case characterized by T2k > 0, see equation
(25).
An important value of χk we must take into account is the one over which sector 2 is
making negative profits. Let us call it χk, defined as
χk =ω1k
ω2k
=w1 − (1 + γ1k)
N2
N1T2k
w2 + (1 − θ2k)T2k
=
(
N2
N1
)
µ (N1ψkw1 +N2A2) − F
(1 − µ)ψkF(26)
where the last part of the equation follows from inserting the value of T2k given by
equation (25) and rearranging. These cutpoints depend negatively on fixed costs F .
Note also that χA and χB are rather similar, the difference comes from the leaky bucket
parameters associated to each party. In Figure 5 we show these cutpoints and the regions
21
χB
χA
χB
Π2A > 0
Π2B > 0
Π2A > 0
Π2B < 0
Π2A < 0
Π2B < 0
Π2A < 0
Π2B > 0
χA
Figure 5: Firm’s profitability given transfers
of sector 2 profits they define.
Distribution technologies. Both parties have in principle the possibility to either
channelling the transfers to sector 2 workers through the firm or, instead, in a direct
way. Direct transfers are those that bypass industry 2, that is they directly reach
workers without having an impact on firm’s profits. Under Assumption 5, direct transfers
therefore imply sector 2 is shut down, and as a consequence, w2 = 0. The total amount
available to candidate k for redistribution (recall equation 22), would fall to N1ψkw1.
Indirect transfers, correspondingly, are those that help improve the firm’s profits, thus
allowing it to survive, and yielding w2 > 0.
We will use the following notation: For any variable z, we use z if transfers to workers
are indirect; z if direct transfers; and z if transfers are such that sector 2 profits are just
equal to zero.
We consider hereafter the following case: Industry 2 members are a core support
group of party A. This means party A can more easily get political support from
this group using transfers. In terms of the model, the parameter θ2A is close to zero.
Candidate A can be therefore thought of as a professional politician who has been
22
“investing” her time in the sector, and as a result she has a better understanding of
their members. A natural consequence of this closeness is party A is ready to intervene
should this industry go into problems by contributing to its survival through indirect
transfers (which, as seen above, improve the firm’s profitability).
Party A has nevertheless the possibility to let the firm die and make direct transfers.
Indeed, the support to sector 2 has a limit: eventually, the total vote it receives falls
below the votes it would obtain by proposing direct transfers and letting the firm shut
down. When this happens, party A abandons indirect transfers and starts proposing
direct ones. However, this has a huge cost in terms of the candidate’s reputation and
therefore on its political influence over sector 2 workers. In the event the firm disappears,
candidate A would be blamed as responsible. Thus, abandoning indirect transfers means
losing the bulk of its ability to get to voters: the parameter θ2A would increase, resulting
in ψA that falls below ψA. As shown in the analysis of Figure 2, this implies party
A’s best response function moves outwards. The rationale for party A to use indirect
transfers is then given by the fact that it is the best it can do taking into account its
abilities to redistribute.
In turn, candidate B has no such abilities to get to industry 2 members. Think of
him as the outsider, the “reformer” or simply the “new technocrat just arrived from the
capital”. He does not have the close relationship with sector 2 workers candidate A has,
implying in turn he has no incentives to artificially prolong the survival of industry 2.
That is why party B will always propose direct transfers to sector 2 group members.
Notice this strategy implies the shut down of this industry.
3.1 Parties best response functions
The best response functions of both parties in the unconstrained region are those of
section 2, modified to account for the assumptions we have introduced in this section:
χA =
[
δ1
δ2
1
ψA
]1
1−σ
[
1
χB
]σ
1−σ
(27)
23
χB =
[
δ1
δ2
1
ψB
]1
1+σ
χσ
1+σ
A (28)
It is convenient here to define the level of χB that produces a best response χA equal to
χA(F ), i.e. where party A transfer to sector 2 is such that firm’s profits are zero: Let
χB,0 be such that χA(χB,0) = χA(F ). Using equation (27) for χA we have
χB,0 =
(
δ1
δ2
1
ψA
)1
σ
(
1
χA(F )
)1−σ
σ
.
It follows that χB > χB,0 ⇔ χA(χB) < χA and this implies Π2A > 0. Next, let us define
the threshold χB,1 below which the total vote of A under direct transfers is greater than
the vote under T2A, i.e. let χB,1 be such that VA = VA. Applying this definition, χB,1 is
implicitly defined by the following equation:
(
1
χB,1
)σ
N1δ1 [ωσ1A(χB,1) − ωσ1A] −N2δ2 [ωσ2A − ωσ2A(χB,1)] = 0.
where we make explicit the dependence of ωiA, the after tax income promised by party
A, on party B best response function. Therefore, party A strategy can be summarized
by the following best response function:
χA(χB) =
χA(χB) if χB ∈ (χB,0,∞)
χA(F ) if χB ∈ (χB,1, χB,0)
χA(χB) if χB ∈ (0, χB,1)
For levels of χB ≥ χB,0 party A is on the unrestricted area so the best response function
is given by equation (27). For intermediate levels of χB, candidate A best response is
χA(F ) through which it guarantees a level of transfer T2A such that sector 2 profits are
zero, but sector 2 remains open. Finally, for the lowest values of χB, party A abandons
indirect transfers, industry 2 is closed, and direct transfers prevail. Party B, in turn,
always proposes direct transfers to group 2 members. Its best response functions is given
by equation (28).
Parties’ best response functions are shown in Figure 6. In the case represented there,
the equilibrium is found at the intersection of χA(F ) and χB(χA).
24
χB
χA
χA(χB)χB,1
χA(χB)
χB,0
χA(F )
χB(χA)
Figure 6: Parties’ strategies
3.1.1 Equilibria
Given the partitioned best response function of party A, we may have several equilibria.
Moreover, there is a region in which party A has no best response function at all,
implying no equilibrium is defined. The precise configuration of an equilibrium depends
on parameter levels (such as the ideology distribution and the relative abilities of parties
to transfer) and particularly, on the level of fixed costs which positively affects both χB,0
and χB,1. We next explore these equilibria, according to the level of F .
Given that we want to identify the implemented platform in each configuration, we
are interested on finding which candidate wins the elections. For that, we have to feed
the model with some particular assumptions on parameters. Let us start out with the
following ranking of candidate’s abilities to redistribute: ψA > ψB > ψA. Accordingly to
what we have said before, party A has the highest ability when using indirect transfers,
but if direct transfers prevail it loses reputation vis a vis its electorate, then party B has
the highest ability.
25
Case 1 - low fixed costs. When the level of fixed costs is low enough the thresholds
χB,0 and χB,1 are also low, so that the equilibrium is most likely found at the intersection
of χA(χB) and χB(χA). This is the unconstrained case, for which we adapt the standard
model of Section 2, as explained in Subsection 3.1. The intersection of both best response
functions yields the equilibrium pair (χ∗A, χ
∗B), with similar characteristics as those of
equations (14) and (15). However, since party B proposes direct transfers this implies
w2 = 0 and thus the amount available for redistribution is only YB = N1ψBw1. This
implies Proposition 4, which identifies the winner of elections in the standard case, fails
to hold.
For party A to win it must be the case that VA > N2. The total vote for party A
is given by equation (7) that formulates VA as a function of thresholds xi. Together
with the logarithmic utility function, the assumption on the distribution of ideology,
and taking into account the relationship between x∗1 and x∗2 stated by equation (23),
candidate A wins the elections if and only if
x∗2 > x2
where
x∗2 =
(
N1ψAw1 +N2w2
N1ψBw1
)σ
=
(
YA
YB
)σ
,
and
x2 =N1δ1ψ
σA +N2δ2ψ
σA
N1δ1ψσB +N2δ2ψ
σA
.
Then, the condition writes
VA >N
2⇔
YA
YB> (x2)
1
σ .
Proposition 5 In the case where the equilibrium is found at the intersection of the best
response functions χA(χB) and χB(χA), if ψA > ψB then candidate A wins the elections.
Proof. Let us analize how both sides of the inequality behave as functions of N2 and
given the leaky bucket parameters embedded into ψk.
limN2→0
(
YA
YB
)
=ψA
ψB
26
∂(
YA
YB
)
∂N2
> 0
limN2→0
[
x1
σ
2
]
=ψA
ψB
∂[
x1
σ
2
]
∂N2
< 0 if ψA > ψB
i.e. both sides of the inequality are equal when N2 = 0 and, as N2 takes positive
values, the left hand side grows (for any ψA, ψB) whereas the right hand side decreases
if ψA > ψB. Therefore, the condition under which A wins is satisfied for all N2 > 0
provided that party A has more ability to transfer than party B. Note when N2 = 0
there is a tie in elections. 2
Case 2 - intermediate fixed costs. For some level of fixed costs, Party A enters into
the constrained zone, so its best response function is χA(F ), through which it guarantees
that sector 2 can be open and its workers get their salary. The equilibrium is found at
the intersection of this best response function with χB(χA). This is the case represented
in Figure 6. In order to find the winner, note this equilibrium may be replicated by
using a generic best response function of party A associated with a particular level of
ability to transfer, that we will denote as ψfA. Recall how these functions are modified
when ψk changes, see Figure 2. Therefore, the equilibrium in the constrained zone may
be characterized by the intersection of χB(χA) and χA(ψfA), where ψfA > ψA. Therefore,
we can still use Proposition 5 and, since ψfA > ψB, party A wins the elections.
Case 3 - high fixed costs. Finally, if fixed costs are high, party A cannot continue to
propose indirect transfers and it switches to direct transfers. This has a cost in terms of
reputation for this party, that translates into a higher cost of redistribution, modelled
through an increase in leaky bucket parameters. Therefore, the equilibrium is found at
the intersection of χA(χB) and χB(χA) and it entails the shut down of sector 2 (whoever
wins elections). Then, w2 = 0 and we have Yk = N1ψkw1, k = {A,B}, so we can still
use the results of Proposition 4. Thus, under our assumptions i.e. ψB > ψA, party B
wins.
27
We have shown that, for this particular case of parties abilities we are considering, the
inefficient sector is allowed to survive. This is possible because the party which proposes
indirect transfers wins the elections, and this is so provided fixed costs are not to high.
Note we could interpret these fixed costs in a broader manner, say, as reflecting exogenous
variables such as perturbations or other real conditions affecting the economy. Being
allowed to do that, the situation of low and intermediate fixed costs would reflect normal
economic conditions, whereas a high level would imply bad times. This replicates the
experience of many industries and sectors around the world: some of them are born and
survive without any kind of external assistance, others, in contrast, need an increasingly
amount of support from the government. This may initially count with popular support,
given the importance people attach to some industries. Eventually, years of decline and
the arrival of new generations may make people and politicians change their minds, and
they are shut down. Again, the example of the US automobile industry can be used to
illustrate this.
4 Conclusions
The support to a declining industry may have a political rationale. Politicians with a
strong electoral support in some areas or sectors may have incentives to prolonge the
survival of firms that otherwise would not stand competition. This paper has shown
that such a strategy may be a succesful one. When times are good (represented by low
fixed costs) and the cost of keeping the inefficient sector alive is low, the candidate who
is closer to the sector can easily win elections. That may be of no surprise. However, the
paper also shows what happens when times become tougher and the alternative to let
the firm die is viewed as increasingly feasible in terms of votes. Eventually, there is no
politician who is willing to maintain this artificial situation, so that when fixed costs are
high enough both candidates decide to abandon the transfers that help the inefficient
sector to survive.
In this sense, this paper shed some light on the debate about the convenience of
rescuing and giving financial support to in-trouble firms. Moreover, it can be seen as a
28
representation of two opposite visions of political intervention in economic activity: on
the one hand, the (perhaps older) approach of parties promising transfers based on purely
electoral basis (machine politics). On the other hand, many contemporaneous politicians
express, at least in their speeches, their commitment to policies that encourage economic
efficiency.
Related to this, we have the traditional debate on two opposite viewpoints: should
governments grant subsidies to inefficient firms as a way to maintain social stability,
employment or other related target? Or should they let those firms die, allowing fac-
tors to relocate to more productive activities while compensating the losers from reform?
29
5 Appendix
Proof of Proposition 1. We analyse which are the features that an interest group
must have in order to be benefitted from redistribution. The problem of, say, party A
is the following
max{TiA} VA =∑i=2
i=1NiΦi(xi)
s.t.∑i=2
i=1NiTiA ≤ 0 BCA
(29)
The Lagrangean associated to this problem writes:
L =i=2∑
i=1
NiΦi(xi) − λA
i=2∑
i=1
NiTiA (30)
The Lagrange multiplier λA measures the value to party A of an additional unit available
for redistribution, in terms of votes. The first order conditions of this problem are:
∂L
∂TiA= Ni
[
δiσωσ−1
iA
ωσiB
∂tiA
∂TiA− λiA
]
= 0, i = 1, 2. (31)
To see which group is benefitted from redistribution we start out with these first order
conditions, evaluating them at T1A = T2A = 0, for a given platform TiB of party B. The
marginal rate of return of an extra unit available for redistribution is given by
λ1
A(0) = δ1σwσ−1
1
ωσ1B
∂t1A
∂T1A
λ2A(0) = δ2σ
wσ−1
2
ωσ2B
∂t2A
∂T2A
If λ1A(0) > λ2
A(0) party A can do better by distributing more to group 1, this means
T1A > 0. Given the budget constraint, equation (2), this implies T2A < 0. Recalling
equation (1), leaky bucket parameters are given in this case by {θ1A, γ2A} so that
∂tiA
∂TiA=
(1 − θ1A) for i = 1
(1 + γ2A) for i = 2(32)
The condition in this case is
λ1
A(0) > λ2
A(0) ⇔ χB <
[
w2
w1
]1−σ
σ
[
δ1
δ2
1 − θ1A
1 + γ2A
]1
σ
≡ A1.
30
Inversely, if λ1A(0) < λ2
A(0) party A will tax group 1 to subsidize group 2. The condition
for this is
λ1A(0) < λ2
A(0) ⇔ χB >
[
w2
w1
]1−σ
σ
[
δ1
δ2
1 + γ1A
1 − θ2A
]1
σ
≡ A2.
Given the fact that leaky bucket parameters lie between 0 and 1, it is the case that
A1 ≤ A2. For given party B platform, group i will benefit from party A’s redistribution
scheme if its wage wi is low compared to the other group’s wage, and if it is more dense,
i.e. if δi is high. Note there is an intermediate region in which party A does not make
any redistribution, provided at least one leaky bucket parameter is different from zero.
The analysis for party B is analogous, yielding the following condition under which
party B decides to tax group 2 and to subsidize group 1:
χA >
[
w1
w2
]1+σ
σ
[
δ2
δ1
1 + γ2B
1 − θ1B
]1
σ
≡ B1.
And B will tax group 1 to subsidize group 2 whenever:
χA <
[
w1
w2
]1+σ
σ
[
δ2
δ1
1 − θ2B
1 + γ1B
]1
σ
≡ B2.
Note that B1 ≥ B2. The above 4 conditions define the regions we show in Figure 7.
Our equilibrium results of subsection 2.2 must verify the conditions to have T2k > 0
and T1k < 0, k = {A,B}, that is χ∗A < B2 and χ∗
B > A2. Graphically, this equilibrium
must be in the northwest quadrant. Then, recalling equation (15), party A taxes group
1 to subsidize group 2 if and only if:
χ∗B > A2 ⇔
w1
w2
>δ1
δ2
[
1
ψA
]1+σ
ψσB
Similarly, given equation (14), party B taxes group 1 to subsidize group 2 if and only
if:
χ∗A < B2 ⇔
w1
w2
>δ1
δ2
[
1
ψA
]σ [1
ψB
]1−σ
We may now perform the same analysis for the remaining 3 quadrants. In the
southeast quadrant, both parties tax group 2 to subsidize group 1. First let us find the
equilibrium to then check under which conditions it verifies χ∗A > B1 and χ∗
B < A1.
31
χB
χA
A2A1
T2A > 0
T2B > 0
T2A < 0
T2B > 0
T2A > 0
T2B < 0
T2A < 0
T2B < 0
B2 B1
Figure 7: regions
From the first order conditions, equation (31), we find parties’ best response func-
tions:
χA =
[
δ1
δ2ǫA
]1
1−σ
[
1
χB
]σ
1−σ
χB =
[
δ1
δ2ǫB
]1
1+σ
[χA]σ
1+σ
And the Nash equilibrium:
χ∗A =
δ1
δ2
[
1
ǫB
]σ
[ǫA]1+σ
χ∗B =
δ1
δ2[ǫB]1−σ [ǫA]σ
Finally, it must be the case that this equilibrium verifies the following conditions:
χ∗A > B1 ⇔
w1
w2
<δ1
δ2[ǫB]1−σ [ǫA]σ
χ∗B < A1 ⇔
w1
w2
<δ1
δ2
[
1
ǫB
]σ
[ǫA]1+σ
32
It remains to be analyzed the southwest and northeast quadrants, in which parties
differ in strategies. In the northeast quadrant party A taxes group 1 while party B taxes
group 2. The inverse case is found in the southeast quadrant.
The first order conditions, equation (31), for the case T2A > 0, T2B < 0 (northeast
quadrant) yield the following best response functions:
χA =
[
δ1
δ2
1
ψA
]1
1−σ
[
1
χB
]σ
1−σ
χB =
[
δ1
δ2ǫB
]1
1+σ
[χA]σ
1+σ
And the Nash equilibrium:
χ∗A =
δ1
δ2
[
1
ǫB
]σ [1
ψA
]1+σ
χ∗B =
δ1
δ2[ǫB]1−σ
[
1
ψA
]σ
This equilibrium must verify
χ∗A > B1 ⇔
w1
w2
<δ1
δ2[ǫB]1−σ
[
1
ψA
]σ
χ∗B > A2 ⇔
w1
w2
>δ1
δ2
[
1
ǫB
]σ [1
ψA
]1+σ
These two regions define a non-empty solution of w1
w2only if
(1 − θ1B)(1 − θ1A) > (1 + γ2B)(1 + γ2A)
and given that every leaky-bucket parameter lies between 0 and 1, this never happens.
Then this case is discarded. Similar results are found for the last case, corresponding
to the southwest quadrant, in which party A taxes group 1 and party B taxes group
2. As a result, in the political game with leaky-bucket assumption in which two parties
33
compete for votes, there are two strategies characterized by the fact that both parties
decide to tax the same group.
Proof of Proposition 4. We derive the conditions to have party B as the winner (the
procedure to show party A wins is analogous). This is so whenever this candidate gets
more than half of the votes of the population, VB >N2, i.e. when
N1 [1 − Φ1(x∗1)] +N2 [1 − Φ2(x
∗2)] >
N
2
substituting the distribution function given by equation (6) we have
⇔ N1δ1 (x∗1 − 1) +N2δ2 (x∗2 − 1) < 0
inserting the relationship between thresholds given by equation (23) we get a condition
on x∗2
⇔ x∗2 <N1δ1ψ
σA +N2δ2ψ
σA
N1δ1ψσB +N2δ2ψ
σA
Now, we use the equilibrium value of the cutpoint of x2 given by equation (2.3),
⇔
[
N1ψAw1 +N2w2
N1ψBw1 +N2w2
]σ
<N1δ1ψ
σA +N2δ2ψ
σA
N1δ1ψσB +N2δ2ψ
σA
Let us denote the right hand side of the above equation by x2. Consider the case
ψB > ψA. Then,
⇔ N1w1
[
ψB (x2)1
σ − ψA
]
> N2w2
[
1 − (x2)1
σ
]
⇔N1w1
N2w2
>1 − (x2)
1
σ
ψB (x2)1
σ − ψA.
Now, recall the condition to have both parties taxing group 1 to subsidize group 2,
equation 9. Under ψB > ψA, this is the case if
N1w1
N2w2
>N1δ1ψ
σB
N2δ2ψ1+σA
We next show that when partyB has more ability to transfer than party A (i.e. ψB > ψA)
and if both parties tax group 1 to subsidize group 2, then party B wins the elections.
This is so whenever:N1δ1ψ
σB
N2δ2ψ1+σA
>1 − (x2)
1
σ
ψB (x2)1
σ − ψA
34
⇔ (x2)1
σ >ψA (N1δ1ψ
σB +N2δ2ψ
σA)
N1δ1ψ1+σB +N2δ2ψ
1+σA
Inserting the value of (x2) and rearranging,
⇔ N1δ1 +N2δ2 >(N1δ1ψ
σB +N2δ2ψ
σA)1+σ
(
N1δ1ψ1+σB +N2δ2ψ
1+σA
)σ
Let us call the right hand side of this condition Γ(ψB, ψA, δi, σ, Ni). Note
(i). Γ(0, ψA, δi, σ, Ni) = N2δ2
(ii). ∂Γ
∂ψB
< 0 if ψB > ψA.
Then, the condition holds if ψB > ψA. Therefore, in the case in which both parties
tax group 1 to subsidize group 2, if ψB > ψA then party B wins the elections. Under
ψA > ψB, party A wins the elections. 2
35
References
[1] Dixit, Avinash and John Londregan (1996). “The determinants of success of special
interest in redistributive politics”. The Journal of Politics 58, no. 4, 1132-1155.
[2] Dixit, Avinash and John Londregan (1995). “Redistributive Politics and Economic
Efficiency”. The American Political Science Review, Vol. 89, No. 4, December, pp
856-866.
[3] Kornai, Janos; Eric Maskin and Gerard Roland (2003). “Understanding the Soft
Budget Constraint”. Journal of Economic Literature, Volume 41, No. 4, December,
pp 1095-1136(42).
[4] Robinson, James A. and Ragnar Torvik (2005). “A Political Economy Model of the
Soft Budget Constraint”. CEPR Discussion Paper No. 4274, October.
36